[gpaw-users] problem of electronic convergence

[Vladislav Ivanistsev]

Hello,
Application of LCAO was successful. However, when restarting from .gpw file
(h=0.16)
  sys,calc = restart('biemimbf4-LDA.gpw',
              h=0.16,
              mode='fd',
              txt=biemimbf4.txt',
              parallel={'sl_auto':True},
              nbands=272, #Number of Valence Electrons: 436
              basis={'H': 'dzp', 'B': 'dzp', 'C': 'dzp', 'N': 'dzp', 'F':
'dzp'},
              occupations=FermiDirac(0.1),
              poissonsolver=PoissonSolver(eps=1e-12),
              eigensolver='cg',
              xc='LDA',
           )
in fd-mode the energy lowers unreasonably:
                     log10-error:    Total        Iterations:
           Time      WFS    Density  Energy       Fermi  Poisson
iter:   1  08:58:26  +2.8            -231.571270  2      208
iter:   2  09:12:53  +2.8            -230.877292  2
iter:   3  09:27:21  +2.9            -230.751393  2
iter:   4  09:42:47  +2.9   -0.9     -233.512597  8      98
iter:   5  09:58:11  +3.0   -0.9     -237.222281  4      95
iter:   6  10:13:25  +3.0   -0.9     -238.529457  2      72
iter:   7  10:28:45  +3.0   -0.9     -240.845417  3      80
iter:   8  10:44:02  +3.0   -0.9     -240.093878  3      67
iter:   9  10:59:36  +3.0   -0.9     -255.309288  5      96
iter:  10  11:15:08  +3.0   -0.9     -273.295907  3      93
iter:  11  11:30:49  +3.1   -0.9     -339.679072  3      114
iter:  12  11:46:34  +3.1   -1.0     -531.732518  10     117
iter:  13  12:02:17  +3.2   -1.0     -918.309682  11     118
iter:  14  12:18:05  +3.3   -1.1     -1522.414840  19     123
iter:  15  12:33:42  +3.4   -1.2     -1863.915047  15     113

I'm asking for an advice
Many thanks,
Vladislav


On Mon, Jan 28, 2013 at 5:17 PM, Ask Hjorth Larsen <asklarsen at gmail.com>wrote:

> Hi
>
> 2013/1/28 Vladislav Ivanistsev <olunet at gmail.com>:
> > Dear Ask,
> > Is there a way to restart with a .gpw and change grid spacing from 0.24
> to
> > some other value?
> > I've tried to, but GPAW stops notifying about mismatched gpts.
> > Many thanks
>
> I'm afraid that would throw away the wavefunctions and density
> entirely.  In LCAO mode one could save/reload the coefficients by
> doing some code light/moderate hacking, depending on how important it
> is, but it tends to be less important in LCAO calculations as they
> don't last that long, so probably it isn't worth it.
>
> Regards
> Ask
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20130129/23c43e8d/attachment.html